Bet on AI. Bet Against the Hype.

At CES last week, a startup called Rabbit unveiled the R1: a cute orange gadget, $199, powered by what they call a “Large Action Model” that will use your apps for you. It sold out its first batches within days. Another startup, Humane, is about to ship a $699 AI pin you wear on your shirt instead of carrying a phone. Every pitch deck I’ve seen this month has “AI” in the first sentence. A company that makes toothbrushes announced an AI toothbrush.

At the same time, every developer I know is using AI every single day, and quietly getting a lot more done.

Both of these things are true. And I think the most useful thing you can do in 2024 is hold both of them in your head at once.

AI is real

Let me be clear, because people sometimes read me as an AI skeptic. I’m not. I think this is the biggest shift in how software gets built since the web itself, and possibly bigger.

A year ago I called ChatGPT a junior developer with infinite confidence. Since then, GPT-4 arrived, and the junior got noticeably less junior. Today I use AI to draft code, explain unfamiliar codebases, write tests, translate between languages, debug weird errors, and think through designs. It’s not occasionally useful. It’s useful dozens of times a day.

Every month the tools get better. Every month, things that were impossible become routine. If you’re a developer and you’re not using these tools seriously yet, you’re working with one hand tied behind your back, and that gap is going to get wider, not narrower.

Bet on AI. It’s the right bet.

The hype is mostly wrong

And yet. Most of what’s being sold as AI right now is going to disappoint, for predictable reasons:

• Hardware for a software problem. The magic is in the model, which runs in the cloud and is available on the phone you already have. A new gadget mostly adds a worse screen and a monthly subscription.
• Demos instead of products. A demo needs to work once, on stage. A product needs to work a million times, for strangers, on weird inputs. The distance between those two is where most AI startups will die.
• Wrappers. A nice UI over someone else’s model, with no data, no workflow, and no moat. If the model provider can ship your product as a feature, they will.
• “AI will replace X” headlines. Usually written by people who don’t know what X involves day to day.

Bet against the hype. Not against the technology - against the specific claims, timelines, and products that sound too good, because they usually are.

How to tell them apart

My rough test for any AI product or claim:

1. Does it make an existing, real workflow meaningfully faster or better? (Coding assistants: yes. AI toothbrush: I’m going to say no.)
2. Does it still work when the input is messy? Try it yourself, on your real data, not the demo.
3. Who checks the output, and do they know enough to catch mistakes?
4. Would it still be valuable if the word “AI” weren’t in the pitch?

What I’m doing about it

Personally, I’m going all-in on AI as a tool - learning it deeply, building it into how I work, building things with it. And I’m staying skeptical of every specific promise until I’ve seen it work on something real.

That’s not a contradiction. It’s what experience feels like. I saw the dot-com bubble - the hype was absurd, and the Internet still changed everything. Both were true then, too. The people who did well were the ones who believed in the technology and ignored the circus.

The technology is real. The circus is temporary. Don’t confuse the two.